Perceived labor market competition and out-group prejudice: Experimental evidence from refugee and host populations in Malaysia

Our study consists of a randomized questionnaire module that will be implemented as part of a large representative survey. The questionnaire will be conducted with registered refugees and their neighboring host communities across four states in Malaysia, exploring facets such as living conditions, labor market characteristics, and interactions between refugees and hosts.
We are targeting refugee populations and host community members living in close proximity to each other in the regions of Kuala Lumpur, Selangor, Penang and Johor in Malaysia.

Our survey sample will cover approximately 2,800 refugee households (including 1,000 Rohingya refugee households, 1,000 non-Rohingya refugee households from Myanmar, 800 refugee households from other countries) and around 1,400 host community households across 34 mukims (districts) in 4 states of Malaysia.

As part of this survey, respondents will be randomly exposed to different narrative vignettes about a fictitious member of the in-group or two different out-groups with identical or different labor characteristics.

In the case where participants are assigned to an out-group narrative, we distinguish between "close" and "distant" out-groups. For respondents in the refugee group, they are presented with narratives featuring either a refugee from the same identity (in-group), a refugee from a different identity (out-group A, close), or a Malaysian individual (out-group B, distant). Conversely, respondents in the Malaysian (local hosts) group will encounter narratives about a fictitious individual who is either another Malaysian (in-group), a refugee born in Malaysia (out-group A, close), or a refugee born in another country (out-group B, distant). This grouping approach allows us to delve into a unique aspect of the Malaysian context. Given the protracted refugee situation without easy pathways to legal residency or employment rights in Malaysia, there exists a notable number of individuals who were born in Malaysia but still carry a "refugee" status as second- or third-generation refugees. For refugees, our focus lies in investigating biases within the diverse refugee community and gaining insights into their perceptions of individuals who, like them, possess refugee status but belong to different nationalities.
The second dimension of randomization within the vignette relates to the fictitious character’s labor market characteristics, such as their occupation. Our experiment randomly varies this information across respondents so that the occupation either matches or is different from the main occupation exerted by the respondent’s household.

The study evaluates the interplay between perceived labor market competition and out-group discrimination. Based on the outcomes of our experiment, we aim to analyze the prevalent attitudes towards the out-group among hosts and refugees in Malaysia, and assess the extent to which these attitudes are influenced by different concepts of the “out-group”. Moreover, we will ask whether perceived labor market competition is an important driver of discriminatory attitudes towards the out-group.

The intervention in Malaysia is planned to start on November, 8th 2023 and end on March, 31st 2024.

Our experiment evaluates the link between perceived labor market competition and out-group discrimination, using a vignette experiment.

Vignette experiments have been used by previous studies to quantify pre-existing beliefs of the out-group or to modify negative perceptions of the out-group. Some of these studies use short stories or perspective-taking to trigger empathy with vulnerable groups (Adida et al., 2018 ; Cattaneo and Grieco, 2021 ; Rodríguez Chatruc and Rozo, 2021 ). In some cases, the vignette is intended to remind or inform respondents of individual traits that the respondent shares with the out-group, such as belonging to the same religious sect (Lazarev and Sharma, 2017 ) or sharing a family history of forced displacement (Dinas, Fouka and Schläpfer, 2021 ). Vignettes have also been used to expose survey respondents with more accurate information about migrants (Blinder and Schaffner, 2020 ; Facchini, Margalit, and Nakata, 2022 ; Grigorieff, Roth and Ubfal, 2020 ; Haaland and Roth, 2020 ; Hopkins, Sides and Citrin, 2019 ). Such interventions have been shown to improve the respondents’ willingness to engage with the out-group, but only along certain social dimensions, and the impacts may be short-lived.

We can also relate the experiment to a growing body of papers that look at labor market competition and labor market discrimination between refugees and hosts. Sahin Mencutec and Nashwan (2021) and Fakih and Ibrahim (2016) describe multi-faceted perspectives on labor market integration of Syrian refugees in Jordan. Clemens et al. (2018) refer to key contextual factors influencing refugees’ labor market integration and perceived competition from the side of the local population. Loiacono and Vargas (2019) measure discrimination in hiring decisions towards refugees.

We seek to link these strands of the literature by not only quantifying existing perceived labor market competition and out-group discrimination between local hosts and refugees - but by also analyzing whether high perceived labor market competition is associated with an increase in out-group discrimination.

To do so, we are proposing the following crossing of two experimental treatment arms:
Individuals will listen to a narrative about a fictitious individual who either belongs to the in-group or alternatively a close or distant out-group (for refugee respondents: refugees of the own nationality; of a different nationality; hosts; for host respondents: another host; refugees born in Malaysia; refugees born in another country) that either shared or not the same occupation as the respondent.

Same occupation Different occupation
In-group T1 T2

Out-group A
(close) T3
T4

Out-group B
(far) T5 T6

This 3x2 experimental matrix enables us to disentangle the mechanisms behind prejudicial attitudes against out-group members.

Considering both the horizontal and vertical variation in our treatment arms together will enable us to disentangle whether or not discrimination against the out-group members is, in reality, a fear of loss of livelihood rather than merely a fear “of differences” against out-group members.

Given the matrix of treatment arms, we can test five different hypotheses. We will always conduct the analyses separately for refugee and host respondents, given the different channels at play for each group.

Hypothesis 1: Host communities and refugees hold higher prejudicial views against members of the out-group, compared to members of the in-group.

Comparing pooled measures of prejudice (see section “Primary Outcomes”) for treatment groups T1 and T2 to treatment groups T3, T4, T5, and T6 allows us to quantify discriminatory views of the host population towards refugees and vice versa - independently from their labor market status.

Hypothesis 2a: Host communities hold higher prejudicial views against refugees who are born in another country, than towards refugees who are born in Malaysia.

Among hosts, we will test whether respondents hold stronger prejudicial views towards refugees when they are not born in Malaysia, compared to refugees who are born in Malaysia. We can do so by comparing prejudicial attitudes expressed by Malaysian respondents in T5 and T6 with prejudicial attitudes expressed by Malaysian respondents in T3 and T4.

Hypothesis 2b: Refugees hold higher prejudicial views towards other refugees with a different nationality than towards Malaysian hosts

For refugees, out-group discrimination depends on whether the out-group member is from the hosting population (i.e. Malaysian) or a refugee from another nationality. We hypothesize that there is a “host effect” which countervails potentially negative out-group prejudice: If refugees are treated fairly and consider the host country a safe haven, it could reduce their prejudice towards Malaysians as an out-group in comparison to their prejudice towards refugees of other nationalities. Malaysians are thus not perceived as an “out-group”, but as a “hosting group”.

We expect the opposite effect to be true if there are few integrative efforts, or even social rejections and exclusion. If refugee respondents perceive the Malaysian host community as a threat, they may unite with other refugee groups facing similar challenges, which could result in greater out-group prejudice towards Malaysians compared to other refugee groups with a different identity. We can analyze the channels behind this hypothesis descriptively by examining integration-related questions (i.e. friendships and economic networks with the host population).

We test this hypothesis by comparing the prejudicial views of refugee respondents in treatment groups T3 and T4 versus the attitudes expressed by refugee respondents in treatment groups T5 and T6.

Hypothesis 3: For both refugees and hosts, perceptions of labor market competition may have adverse effects on the views of the other

By comparing pooled measures of discrimination for treatment groups T1, T3, and T5 to treatment groups T2, T4, and T6, we are able to analyse whether higher perceived labor market competition (proxied by shared occupation) is linked to higher levels of prejudicial attitudes.

This effect is a priori unclear: While a shared occupation may result in higher levels of perceived labor market competition and thus stronger negative views of the fictitious individual, it may also enhance the feeling of a shared labor market identity, thereby reducing discriminatory views.

We may disentangle the competition from the identity channel by exploring the heterogeneous effects of hypothesis 4 for respondents who indicate high versus low levels of perceived labor market competition with the fictitious individual, i.e. respondents for which the treatment was more/ less effective in creating a feeling of competition. As a more objective measure, we can rely on household-level and local labor market characteristics (such as the density of the main household occupation within the refugee and host sample at the local level).

Hypothesis 4: For both refugees and hosts, discrimination against members of the out-group is more pronounced when perceptions of labor market competition are strong.

In this central hypothesis, we analyze the interplay between labor market competition and out-group discrimination. We hypothesize that discrimination against the out-group does not happen in a uniform way, but depends on the perceived threat of labor market competition arising from this group.

We thereby test whether there is targeted discrimination where individuals discriminate against the out-group because of their perception of labor market competition. From a conceptual point, this would imply that the difference in attitudes towards the in-group with the same occupation (T1) and the out-group with the same occupation (T3 and T5) is larger than the difference in attitudes towards the in-group with different occupation (T2) and the out-group with different occupation (T4 and T6).

Our premise is that hosts and refugees may feel more threatened if they face direct competition in the labor market, while they do have more welcoming attitudes when they are relatively far in terms of labor characteristics from the out-group member. Discrimination is then triggered mostly by labor market competition and not out-group membership per se.

Hypothesis 5: The mediating effect of labor market competition in increasing discrimination towards the out-group is even stronger in settings where the out-group is more “distant”.
We can extend hypothesis 5 by again differentiating between out-groups A (close) and B (far), to mirror the analysis of Hypothesis 2a and 2b.
In a triple-difference approach, we can analyze whether refugee respondents hold higher prejudicial views towards other refugees of a different identity (compared to Malaysian hosts), whenever perceived labor market competition is high.
We also analyze whether Malaysian respondents are more prejudiciant towards refugees not born in Malaysia (compared to refugees born in Malaysia), whenever perceived labor market competition is high.

Regression approach

Given the different channels at play, and the different framing of the vignette, we will always run separate regressions for refugee and for host respondents. Please note that Hypothesis 1 - 3 can be tested for all respondents, while Hypothesis 4-5 can only be tested for respondents that belong to a household where at least one individual is in the labor force.

In our heterogeneous analysis, we will analyze to which extent out-group prejudice is linked with individual socio-demographic characteristics or spatial properties.

2023-11-08

2024-03-31

Our primary measures of interest are:
(1) Measures of prejudice towards the fictitious individual (based on a prejudice index that can be split into three dimensions: private, social, and work-related interactions)
(2) Behavioral stakes, such as the willingness to actually share the phone number and interact with the character described in the vignette
(3) Discriminatory attitudes with respect to labor market participation of the fictitious individual

Questions belonging to measure (1) capture discriminatory attitudes. They cover questions on whether the respondent would accept an out-group member in their community, family, or as a work colleague. We will use these questions to build an Anderson index of how welcoming the respondent is towards members of the out-/ in-group.

Questions belonging to measure (2) include real behavioral stakes. The survey respondent is asked whether he/ she is fine if their phone number is shared with someone like the fictitious individual and how frequently they would like to interact. Those questions reflect a stronger commitment to interaction with the fictitious character than measure (1) which
captures discrimination in hypothetical situations.

Moreover, the survey includes a set of questions aimed at gauging labor-market-related discrimination (measure 3). We include normative questions eliciting the desired degree of labor market integration for the fictitious individual (should be hired; should start his own business) and the deserved remuneration (more/ less/ the same as the respondent).

These three sets of questions will allow us to analyze
(1) The level of discrimination towards the out-group compared with the in-group, with a special focus on close versus distant out-groups (see hypotheses 1 and 2)
(2) The degree of perceived labor market competition when individuals are sharing or not the same labor market characteristics (see hypothesis 3)
(3) The interplay between the two: Does discrimination increase in situations where perceived labor market competition is high? (See hypotheses 4 and 5)

We will examine the vignette's impact on increased perceptions of competition, when it features a character with the same occupation.

We will additionally analyze heterogeneous effects based on

(1) Socio-demographic characteristics of the respondent (gender, employment status, income level etc.)
(2) Measures of past interactions (friendship, trust, economic and exchange networks) with the in- and out-group
(3) Experienced ethnic discrimination by the respondent
(4) Local characteristics, such as the density of refugees in the area and local labor market characteristics, which can serve as a measure of “actual” competition

In order to analyze whether the vignette indeed induced higher feelings of competition when containing a narrative of somebody sharing the occupation, we regress measures of competition (“I feel in competition with the character.”; “I fear people like him take away my job”) on a treatment dummy of sharing the same occupation with the fictitious character (=1 for T1, T3, T5; =0 for T2, T4, T6).

The heterogeneous analysis will be based on the following measures:

Measure (1) relies on socio-demographic characteristics of the respondent (e.g. gender, employment status, age, income level) that we capture in earlier modules of the questionnaire.

Measure (2) relies on questions from the section on refugee-host interactions (section 10) which is preceding the experimental vignette in section 11. In this section, we capture whether the respondent had friendship links with national Malaysians or refugees, whether his experiences with Malaysians or refugees were pleasant, and the trust level towards Malaysians or refugees.
For both refugees and Malaysian respondents, we can additionally measure integration with the out-group based on economic networks. These comprise financial exchanges (borrowing money) and support with residence and job search as well as health-related assistance.

Measure (3) assesses the frequency with which respondents feel they are treated with less respect due to their ethnicity or nationality. Through a heterogenous analysis of this measure, we aim to determine whether perceived personal discrimination is associated with the level of discriminatory attitudes towards others.

Measure (4) will be based on the local properties of our sample, such as the density of refugees within different “mukims” (Malaysian subdistricts, our PSUs) and local labor market characteristics like unemployment rates or density of specific jobs.

Our experimental intervention is embedded in a survey questionnaire focused on labor market outcomes and interactions between refugees and host communities in Malaysia.

As part of this experiment, respondents are randomly assigned to six equally sized treatment groups. Each respondent will listen to a single narrative about a fictitious individual who is either from the in-group or the out-group (out-group A (close); or out-group B (distant)) and sharing or not the main occupation exerted by the respondent’s household.
T1: In-group; same labor market characteristics (17%)
T2: In-group; different labor market characteristics (17%)
T3: Out-group A; same labor market characteristics (17%)
T4: Out-group A; different labor market characteristics (17%)
T5: Out-group B; same labor market characteristics (17%)
T6: Out-group B; different labor market characteristics (17%)

The experiment (module 11) is implemented towards the end of the survey. Its content is based on information from the labor market module (module 8) as well as the group membership of the respondent, which is gathered in the initial interview questions (S0.3). The experiment is directly preceded by a module on refugee-host relations (module 10), which serves as a balance check of the randomization process and provides a range of heterogeneous dimensions for descriptively testing the contact hypothesis within the experiment.

After the vignette is read out loud to the respondents, they are asked to respond to questions about the fictitious individual, notably concerning perceived labor market competition and general attitudes (prejudice, discrimination).

The purpose of our study is to better understand discriminatory attitudes towards the out- and in-group and to gauge the extent to which negative/positive perceptions of refugees among host communities as well as obstacles/drivers to integration among refugees are driven by the perception that both groups are competing or not in the labor market.

The experiment proceeds as follows:

(1) At the beginning of the survey (module 0), the interviewer classifies whether the respondent is a Malaysian national or a refugee of a particular nationality (based on the UNHCR listing of refugees). This classification is confirmed by the respondent.

Within the detailed survey questionnaire, respondents are asked to report whether any of their household members is currently working and what the main occupation of the household (in terms of livelihood) is.

The wording of the vignette links back to this information, by plugging the string of the respondent’s group membership (“refugee from country C”/ “Malaysian”) and the relevant household occupation into the text of the vignette. Separating the construction of the one-shot experiment from the previously indicated group membership and occupation strings within the survey has the advantage that it reduces experimenter demand effects: respondents are less likely to realize that the information in the narrative purposefully matches (or not) their own characteristics and that they are participating in a randomized module.

(2) Respondents are randomly assigned to any of the six equally sized treatment groups (T1 - T6) as described above. Randomization is executed by a computer right before the implementation of the experimental module (11), to avoid unbalancedness due to attrition.
Due to the matching process along with labor market characteristics, only those respondents who indicate that at least one household member is currently working (and thus indicate a main household occupation) can be assigned to T1, T3 or T5.


(3) In the experimental section, respondents will be randomly exposed to a standard narrative about a fictitious individual who belongs to group G and is exerting occupation O.

The random narrative stresses the occupational background of the fictitious character, as well as his identity as a being local Malaysian or refugee of a particular nationality (for Malaysian respondents assigned to an “out-group” treatment, it stresses instead whether the fictitious refugee is born or not born in Malaysia). Narratives are only differing with respect to single words that are inserted for occupation O and group G.

O = {main occupation of the respondent’s household; random occupation R}
The randomly chosen occupation is drawn from a list of jobs that can be plausibly exerted by refugee and host individuals.: R= {Construction laborer, sweeper, cleaner, farmer, restaurant waiter, cook, hawker, salesperson, driver, delivery rider, mechanic, machine operator}

For refugee respondents:
G = {Malaysian; refugee from country X} where X={Rohingya from Myanmar; Burmese from Myanmar; Pakistan; Yemen; Syria; Iraq; Bangladesh; Somalia; Afghanistan}
For host respondents:
G = {Malaysian; refugee from another country born in Malaysia; refugee born in another country}

Single words for group G and occupation O will be auto-filled by a computer, based on survey response in the modules preceding the experimental section.

T1: Respondent listens to a narrative about an in-group member who is working in the same occupation as his household.
T2: Respondent listens to a narrative about an in-group member who is working in a different occupation than his household.
T3: Respondent listens to a narrative about an out-group member (close) who is working in the same occupation as his household.
T4: Respondent listens to a narrative about an out-group member (close) who is working in a different occupation than his household.
T5: Respondent listens to a narrative about an out-group member (far) who is working in the same occupation as his household.
T6: Respondent listens to a narrative about an out-group member (far) who is working in a different occupation than his household.

The narratives take the following form:

“Adam is [group G: same group as respondent’s/ different from respondent’s]. He has (lived in Malaysia his entire life and) moved to the district of [match district of respondent] five years ago. He has been working as a [occupation O: same occupation as respondent’s household/ different occupation] for a long time, so he has a lot of experience in his occupation. He also speaks Bahasa and English very well. He enjoys working in this profession and would recommend his friends to work in the same sector. But while being a [occupation O: same occupation as respondent’s household/ different occupation] fulfills him, he is sometimes very tired after work. Due to difficult circumstances, he has to change jobs while keeping his current profession. So far, he has struggled to find a job.”

The narrative stresses competition in the local labor market by referring to the fictitious character as a local neighbor, living in the same district/ capital. Stories are comparable across all groups (refugees; hosts; different occupations): all four treatment groups’ narratives have the same length and are containing identical information (e.g. duration of stay in this district - thereby similar knowledge and networks; language skills; challenge in finding employment).

The narrative is designed to be politically benign and to avoid raising positive or negative emotions. The vignette always alludes to a fictitious male individual, since the female employment rate is comparatively low in Malaysia, even more so among the refugee population.

While inserting a string for group level G is straightforward, we are implementing one additional step when it comes to occupation group O to make narratives exactly comparable in their length (one to three words).

For individuals who have been randomly assigned to treatment groups T1, T3, or T5 (same occupation), the enumerator will be asked to shorten the occupation string indicated by the respondent.
E.g.: If the respondent states he is an “undergraduate teacher for Maths and sports”, the enumerator is trained to shorten the string to “teacher”.
The narrative will then be based on the shortened version of the string.

For individuals who have been randomly assigned to treatment groups T2, T4, or T6 (different occupation), the computer will randomly draw from the list of occupations mentioned above, that can plausibly be exerted by both Malaysians and refugees.

(4) After the fieldworker reads the narrative to the respondent, the respondent answers questions that
- Make them relate to the story (“Should the character search for work? What would you recommend him/ her to do?”)
- Elicit their willingness to interact with the fictitious individual.
- Elicit actual behavioral stakes, like sharing the phone number and actively interacting subsequently
- Gauge their normative perception concerning labor market integration of the fictitious individual

Computer-based during the survey

Individual

No

34 mukims in 4 Malaysian states. The choice is based on random sampling involving Probability proportionate to size.

2800 refugee households and 1400 host community households The experiment is part of a large household survey. While the initial sections on socio-demographics and labor market characteristics collect information for all or at least several randomly chosen individuals (answered by a single respondent on behalf of the other members), the experimental module is administered with a single respondent per household. Consequently, we do not have to cluster our standard errors at the household level and we are avoiding spillover effects. In an ideal scenario, the number of experimental observations will match the number of households in the sample. We are however expecting a certain attrition and non-response rate given the refugee-host context. Since randomization is implemented at the beginning of the experimental module and the sensitivity of the experimental content is generally low and comparable across all six treatment groups, we expect attrition to be random, and proportionally distributed across treatment groups. Another correction we will make to our sample for relevance is that part of our main analysis (Hypothesis 3, 4 and 5) can exclusively rely on individuals whose household is currently working. For these regression analyses, we are excluding households for which all members are out of the labor force, since the aim of triggering different degrees of perceived labor market competition with the vignette is not viable for them. Consequently, the sample size for which we will be able to run this analysis is likely to reduce slightly, especially among the refugee population who is more likely to be out of the labor force.

The whole sample is split into 6 treatment arms. When disregarding attrition and non-response, as well as individuals whose household is out-of-the-labor-force, we are left with 700 respondents per treatment arm.
To account for the distinct perspectives of refugees and hosts, our final analysis will be conducted separately for these two groups, yielding approximately 230 host respondents per treatment group and 470 refugee respondents per treatment group.

The analysis will be conducted separately for refugees and hosts, and will enable us to separately examine five distinct hypotheses. This results in treatment and control groups that differ in size for each of the subsequent hypotheses. Consequently, we have provided below the Minimum Detectable Effect size for refugees and hosts separately across the five hypotheses. Hosts (N=1400) Refugees (N=2800) H1: n1=933 (treatment group) n2=466 (control group) Power=0.8 MDE=0.159 n1=1866 n2=933 Power=0.8 MDE=0.112 H2: n1=466 n2=466 Power=0.8 MDE=0.184 n1=933 n2=933 Power=0.8 MDE=0.130 H3: n1=700 n2=700 Power=0.8 MDE=0.150 n1=1400 n2=1400 Power=0.8 MDE=0.106 H4: n1=466 n2=933 Power=0.8 MDE=0.159 n1=933 n2=1866 Power=0.8 MDE=0.112 H5: n1=233 n2=1166 Power=0.8 MDE=0.201 n1=466 n2=2333 Power=0.8 MDE=0.142 These MDEs are realistic, given the results of similar strands of literature (e.g. Cattaneo and Grieco, 2020 ).